Estimation Of Alpine Grassland Aboveground Biomass And Its Dynamics Variations Characteristics In The Qinghai-Tibet Plateau

As an important indicator of vegetation activities,the aboveground grassland biomass reflects the primary productivity of grassland,which is of great significance to terrestrial ecosystem carbon cycling.It is the basis for quantifying the terrestrial carbon budget to study the temporal and spatial patterns of grassland biomass and its relationship with the environment and it can effectively measure the stability of grassland ecosystem and the sustainable development of grassland ecological resources by accurate estimation the aboveground grassland biomass and its dynamic variations characteristics.The Qinghai-Tibet Plateau has vast grasslands and provides important production resources for pasture animal husbandry,simultaneously,it also serves as the"carbon reservoir","water tower"and other ecological service functions,which reflect very important production and ecological values.Therefore,it is of great significance to the sustainable development of the agriculture and animal husbandry and ecological protection to construct an accurate aboveground grassland biomass inversion model and analysis its temporal and spatial dynamics characteristics and its response to climate in the Qinghai-Tibet Plateau.This research used the Qinghai-Tibet Plateau as the study area,established aboveground grassland biomass inversion model by applying partial least squares,random forest,BP neural network,and deep belief network,with remote sensing data,meteorological data and geographical data.The random forest inversion model was finally selected to estimate the aboveground grasslands biomass In the Qinghai-Tibet Plateau from 2001 to 2019 by analyzing the accuracy and stability of all models,and based on the result,to analyze the temporal and spatial variations trend characteristics of the aboveground grassland biomass and its response to climate factors in the Qinghai-Tibet Plateau by using the Theil-Sen median trend analysis,the Mann-Kendall test,the coefficient of variation,the Hurst exponent and the partial correlation coefficient analysis.The research indicates that:（1）The machine learning models showed excellent performance in this research,and its accuracy and stability were higher than the parameter statistical model and deep learning model.The fitting results of the random forest inversion model（R²=0.84,RMSE=8.51g C/m²,MAE=6.46g C/m²）are more convergent than the BP neural network inversion model,and the random forest regression model reflects higher stability（random forest:R²=0.76,RMSE=9.24g C/m²,MAE=8.30g C/m²）.We conclude that the random forest inversion model is the best biomass estimation model in this study.（2）The fusion of multi-source data showed great potential in biomass estimation.The accuracy of the inversion model based on remote sensing variables were lowest and,as the input variables increase,the accuracy of all models were improved,which indicates that the fusion of multi-source data is one of the important ways to improve the accuracy of grassland biomass inversion models.（3）It showed obvious spatial heterogeneity of the aboveground grassland biomass spatial distribution with a gradual decrease from southeast to northwest distribution and showed a fluctuating increase trend at a rate of 0.1159g C/m²a.Similarly,the aboveground grassland biomass variations trend also showed significant spatial differences,affected by the geographical factors and climatic conditions,for 24.79%of alpine grasslands had an increasing trend,2.32%of alpine grasslands showed a downward trend.（4）During 2001 to 2019,the aboveground grasslands biomass variations of alpine grasslands in the Qinghai-Tibet Plateau were in a moderately stable state(CV_AGB=0.1).50.13%of the aboveground grassland biomass variations were in a state of high stability and relatively high stability,43.77%of the aboveground grassland biomass variations were in a moderately stable state,and 6.1%of the aboveground grasslands biomass variations were in a low and relatively low stability state.（5）In the future,the aboveground grasslands biomass variations of alpine grasslands in the Qinghai-Tibet Plateau will be in a state of weak anti-persistence（Hurst=0.43）.Combined with the results of Theil-Sen median trend analysis,6.22%of the aboveground grassland biomass will increase in the future,while 22.46%of the aboveground grassland biomass will decrease in the future.（6）Temperature and precipitation promote the alpine grasslands vegetation growth in most of the study area.39.1%of the aboveground grassland biomass had a negative partial correlation coefficient with the average temperature in the growing season.60.9%of the aboveground grassland biomass had a positive partial correlation coefficient with the average temperature in the growing season;83.34%of the aboveground grasslands biomass had a positive partial correlation coefficient with the total precipitation in the growing season,and 16.66%of the aboveground grasslands biomass had a negative correlation with the total precipitation in the growing season.